Connection.Questions Make your own

Questions to ask your crush

These are built for a text thread, not a table — short enough to send, open enough to answer with more than one word, and none of them make it obvious that you have been thinking about what to send.

Easy ones, to start

Nobody opens with the hard question. These are the ones you can ask with the plates still on the table, and they still tell you something.

Questions that actually go somewhere

The middle rung. Still safe, but the answer is about who someone is rather than what they did, which is the whole difference.

When something is off

The ones people skip. Skipping them is usually why the same thing keeps coming back.

Where things are going

Fewer of these on purpose. A conversation that is only about the future is a planning meeting.

Some questions are easier to answer when neither of you chose them.

Turn one over. You would not have picked it — that is the point.

Six questions from this page, face down.

On the site they live on a wheel. You spin, it lands, nobody chose it.

Now make one that knows you → $11 once · your photos and inside jokes · yours to keep

Does any of this actually work?

The uncomfortable finding first: text is the medium people choose and the one that underperforms. In a field experiment, people asked to reconnect with an old friend by phone rather than email felt more connected afterwards than those assigned email. In the lab, strangers paired over video, voice or voice-chat formed stronger bonds than those paired over text — and, critically, with no increase in awkwardness. People predicted the voice conditions would be more uncomfortable. They were not. The preference for text is built on an expectation that the data does not support.

So the honest job of a question sent over text is to get to a call, not to substitute for one. That is a smaller claim than this genre usually makes, and it is the one the evidence carries.

The other thing worth knowing is that you cannot screen for this in advance. Two speed-dating studies had participants complete more than 100 self-report measures of traits and preferences before meeting anyone. Machine-learning models predicted 4–18% of how much a person wanted others in general and 7–27% of how much they were wanted — and none at all of whether two specific people would want each other. Every hour spent working out whether they are right for you before you have actually talked is spent on the one quantity that turned out to be unpredictable.

Kumar & Epley (2021), It’s surprisingly nice to hear you: Misunderstanding the impact of communication media can lead to suboptimal choices of how to connect with others, Journal of Experimental Psychology: General · source
Joel, Eastwick & Finkel (2017), Is Romantic Desire Predictable? Machine Learning Applied to Initial Romantic Attraction, Psychological Science · source

No question makes someone interested who isn't. What these do is give a conversation somewhere to go on a night when it would otherwise die at "haha same" — which is a smaller claim than the search results on this topic will make you.

Last checked 28 August 2026.